Rapid development and scale-up of a 1 H -4-substituted imidazole intermediate enabled by chemistry in continuous plug flow reactors

Article abstract of DOI:10.1021/op200351g

The development of reactions in a continuous fashion in plug flow tube reactors (PFR) offers unique advantages to the drug development and scale-up process and can also enable chemistry that would be difficult to perform via batch processing. Herein, we r

Full text of DOI:10.1021/op200351g

Article
pubs.acs.org/OPRD
Rapid Development and Scale-Up of a 1H-4-Substituted Imidazole
Intermediate Enabled by Chemistry in Continuous Plug Flow
Reactors
Scott A. May,* Martin D. Johnson,* Timothy M. Braden, Joel R. Calvin, Brian D. Haeberle, Amy R. Jines,
Richard D. Miller, Edward F. Plocharczyk,‡ Gregory A. Rener, Rachel N. Richey,
Christopher R. Schmid,▼ Radhe K. Vaid, and Hannah Yu
Chemical Product Research and Development, Eli Lilly and Company, Indianapolis, Indiana 46285, United States
S
* Supporting Information
ABSTRACT: The development of reactions in a continuous fashion in plug flow tube reactors (PFR) offers unique advantages
to the drug development and scale-up process and can also enable chemistry that would be difficult to perform via batch
processing. Herein, we report the development of two different continuous flow approaches to a key 1H-4-substituted imidazole
intermediate (5). In a first generation approach, rapid optimization and scale-up of a challenging cyclization reaction was
demonstrated in a PFR under GMP conditions to afford 29 kg of protected product 2. This material was further processed in
batch equipment to deliver di-HCl salt 4. This first generation approach highlights the rapid development of chemistry in
research-scale PFRs and speed to material delivery through linear scale up to a pilot-scale PFR under GMP conditions. In a
second generation effort, a more efficient synthetic route was developed, and PFRs with automated sampling, dilution, and
analytical analysis allowed for rapid and data-rich reaction optimization of both a key cyclization reaction and thermal removal of
a Boc protecting group. This work culminated in 1 kg demonstration runs in a 0.22 L PFR for both continuous steps and shows
the potential of commercialization from a lab hood footprint (1−2 MT/year).
then transformed in two steps to amine 5 via alkylation, protecting
group removal, and neutralization (Scheme 1).
INTRODUCTION
■
The synthesis of nitrogen-containing heterocycles has long
been a topic of intense research.1,2 This is due, in large part, to
the importance of these compounds as drug candidates. The
vast majority of new molecular entities (NMEs) contain at least
one nitrogen atom in the chemical structure. A subcategory of
these compounds are imidazoles, which are notable pharma-
cophores in a number of areas of discovery chemistry research.3,4
Appropriately, numerous synthetic approaches to these compounds
have been published in the literature.5 We required a robust
synthetic route toward 1H-4-substituted imidazole 5 (Scheme 1),
and the existing synthetic approaches to these ring systems, while
well documented, resulted in poor isolated yields of 2 due to the
tendency towards starting material (1) oligomerization under
the reaction conditions. Lower temperature regimes (room tem-
perature →70 °C) favored oligomerization, while at higher
temperatures (>100 °C), cyclization was favored. Lower yield
would be anticipated with increased scale due to less efficient heat
transfer as the batch reactor size increases. To enable rapid scale-up
of this reaction for an early phase delivery, an approach involving a
plug flow reactor (PFR) was used. The high surface area-to-volume
ratio afforded by the PFR allowed for rapid heat-up to a productive
temperature regime with the ability to maintain constant time to
reaction temperature with scale-up. This allowed the key imidazole
cyclization to scale linearly from 1.75-mL research-scale reactors to
a 7.1-L pilot-scale reactor. Additionally, the ability to run at high
pressure in a tubular reactor allowed for the use of a lower-boiling
solvent (MeOH) that greatly simplified the isolation of 2. This
process afforded 29 kg of 2 after batch isolation. Imidazole 2 was
In a second generation approach to 5, the synthetic strategy was
significantly altered to address processing concerns, and N-methyl
ketoamide 6 was used as the starting point (Scheme 2). PFRs
were utilized in both the cyclization of 6 to afford imidazole 3 and
in a thermal protecting group removal to afford free amine 5. This
effort also utilized automated sampling and dilution allowing for
data-rich experimentation, fast reaction optimization and scale up.
Herein we describe the development and continuous processing
technology used in both first- and second generation routes to
1H-4-substituted imidazole intermediate 5.
RESULTS AND DISCUSSION
■
First Generation Approach to Ketoamide 1. The initial
small-scale approach to ketoamide 1 was first developed for the des-
fluoro analogue 11 (Scheme 3). Direct displacement of phenacyl
bromide 76 with hexamethylenetetramine (HMTA) in chlorinated
solvents led to the corresponding amine salt 8, which was isolated
by filtration. Hydrolysis under aqueous acidic conditions led to low
and unreproducible yields of phenacyl amine hydrochloride salt 9.
Phenacyl amine 9 was found to be contaminated with residual
ammonium chloride, which could not be easily removed. This, in
turn, complicated the coupling reactions between 9 and N-Boc-
isonipicotic acid (10) and led to mixtures of the primary amide of
Special Issue: Continuous Processes 2012
Received: December 3, 2011
Published: April 19, 2012
© 2012 American Chemical Society
982
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Scheme 1. First generation approach to N-methyl imidazole 5
Scheme 2. Second generation approach to N-methyl imidazole 5
Scheme 3. Initial synthesis of imidazole 13 via des-fluoro ketoamide 11
Scheme 4. Synthesis of ketoamide 11 via azide/Staudinger sequence
10 along with the desired product (11), which were also difficult
to purify. The synthesis of 12 was then completed by thermal
cyclization with ammonium acetate in 1-butanol, which afforded
1H-imidazole 12. Treatment of the imidazole product directly with
gaseous HCl delivered the amine salt 13 in good yield and purity.
By building off of the general approach shown in Scheme 3,
an alternative route was developed to tosylate salt 15, which
had excellent physical properties and could be obtained in high
yield and purity (Scheme 4). The displacement of phenacyl
bromide 7 with sodium azide in water-wet tetrahydrofuran7 af-
forded azide 14 which was telescoped directly into a Staudinger
reduction.8 Using 1.05 equiv of triphenylphosphine in THF in
the presence of 1.0 equiv of p-toluenesulfonic acid mono-
hydrate the crude azide (14) was efficiently reduced, and the
corresponding tosylate salt 15 crystallized directly from solu-
tion in good yield and in excellent purity. The amide coupling
reaction with 2-propanephosphonic acid anhydride (T3P)
afforded ketoamide 11 in high yield and purity (Table 1).
The chemistry in Scheme 4 was then scaled up and also
applied to desired ketoamide analogue 1 (Table 1). On 22 L
scale the process performed well as the amine salts 15a and 15b
were all isolated in good yield and purity. The coupling reaction
983
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Table 1. Scale-up results for the synthesis of ketoamides 11 and 1
*
Conditions: CDI, EtOAc/THF.
Scheme 5. Small-scale batch synthesis of imidazole 13
between 15 and N-Boc-isonipecotic acid (10) performed well
using T3P in THF containing N-methylmorpholine. However,
due to difficulty sourcing T3P on scale, the conversion of 15b
to 1 was accomplished with carbonyl diimidazole in a mixture
of ethyl acetate and THF (entry 3). This process afforded
comparably pure product, but with lower overall yield.
Development of a First Generation Cyclization Reaction.
The initial small scale cyclization reactions with ketoamide 1
were once again based on encouraging results from the des-
fluoro compound 11 (Scheme 5). Cyclization in butanol in a
sealed tube9 allowed for complete conversion (11→12) in 24 h.
The reaction optimization and workup was straightforward and
amine salt 13 was obtained in 87% yield upon treatment with
gaseous HCl in n-butanol. However, the cyclization reaction
slowed markedly in a scale up attempt in traditional glassware due
to loss of ammonia from the reaction at elevated temperature. This
reaction required multiple charges of ammonium acetate and,
ultimately, resubmission of the crude product to the reaction
conditions to achieve full conversion.
In order to scale this chemistry to kilogram quantities we sought
a more efficient process and turned to a PFR approach where the
cyclization reaction could be conducted at high temperature and
pressure in methanol (Figure 1). Continuous reactions in this paper
required temperatures in the range 100−270 °C, pressures up to 70
bar, and mean residence time (τ) in the range 5−90 min. Tubular
reactors were constructed from 316 L seamless stainless steel which
provided good corrosion resistance with high pressure and tem-
perature ratings at low cost. Nine PFRs with volumes from 1.75 to
7140 mL were used in the chemistry described in this paper. All
nine PFRs used in this study have been characterized (Table 10) on
the basis of axial dispersion number (D/uL), pressure drop, and
heat transfer surface area per unit volume (A/V). It is important to
recognize that actual mean residence time (τ) and calculated mean
Figure 1. Continuous thermal tube reactor setup for reaction
optimization.13
residence time based on reactor volume and feed flow rates are
typically different. In a later section we will show an example where
τ differs by as much as 40% because of density and phase changes at
higher temperatures. In this paper we refer to mean residence time
(τ) only when measured on the basis of F-curve output, which is
described in a later section of this paper. In the absence of a
measured residence time (τ) we report this value as total reactor
volume (Vreactor) divided by the total volumetric flow rate (Qfeed).
Methanol was chosen as solvent since it allowed for homo-
geneous solution feeds of ammonium acetate and ketoamide 11
at room temperature. The crude product (12) was also soluble
in methanol at room temperature which minimized the risk of
reactor fouling. High pressure syringe pumps fed starting
materials through stainless steel tubing with 1.59 mm outside
diameter (o.d.) and 0.56 mm inside diameter (d) into a 1.75 mL
PFR10 with a Vreactor/Qfeed = 45 min.11 The PFR was heated in a
programmable GC oven which allowed for rapid temperature
screening or a temperature program if desired.12 The product
solution exited the GC oven, passed through a heat exchanger,
through a back pressure regulator and arrived at a product
984
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
collection vessel which was at room temperature and atmos-
pheric pressure.
in Figure 1, the exception being differences in reactor dimens-
ions (Table 2). Accordingly, the 26× scale-up in reactor volume
and flow rate from Table 2, entries 1 and 2, resulted in almost
identical results. If the process shown in Table 2, entry 2, were
run continuously (24/7), the 541 mL reactor in a GC oven in a
laboratory hood could process 4.6 kg/day or 32.5 kg/week of
ketoamide 11. One can easily see that, even with modest scale-
up beyond this reactor size, very high throughput could be
obtained even in a standard laboratory hood. This reaction was
Seven different conditions were evaluated14 for the cycliza-
tion of ketoamide 11 in a single day (Figure 2). In the case of
scaled up successfully in three different PFRs based on Vreactor
/
Qfeed data and not τ. This was possible because the same ther-
mal expansion of the single liquid phase and reactor characteristics
(same heat-up time, negligible axial dispersion, and negligible
temperature gradients after heat-up) were maintained in each PFR.
With the impressive results from ketoamide 11 (Table 2), we
sought to apply these conditions to ketoamide 1, which contained
an additional fluorine atom on the aromatic ring. In addition, we
required the differentially protected product 2 (which demanded a
more finely tuned set of reaction conditions) to maximize desired
product while minimizing deprotected amine 16 (Scheme 6). This
small variation in structure proved quite impactful, however, as
initial batch cyclization conditions afforded much lower yields
(50−60%) of product 2. In addition to the desired imidazole 2
and deprotected product 16, significant formations of aniline
analogue 17 and dimer 18 were observed.20
The first efforts to obtain larger quantities of 2 involved
batch reactions in 2 L glass pressure vessels with n-butanol as
solvent at 100 °C.21 The cyclization reaction was performed
four times (200−250 g scale each), and the crude product solu-
tions were combined. A single, labor-intensive workup and
isolation afforded ∼500 g of 2 in 56% yield. In contrast to sub-
strate 11, the reaction with 1 became orange upon stirring at
elevated temperature, and sticky yellow solids formed during
isolation attempts. Fortunately, trituration of the crude product
Figure 2. Temperature screening for the cyclization of 11 in a PFR
with Vreactor/Qfeed = 45 min.
ketoamide 11, loss of the Boc group was not a concern since
the next step was protecting group removal and the
deprotected product (13) could be extracted into an organic
phase along with the desired product 12 during workup. Full
conversion was accomplished with Vreactor/Qfeed = 45 min at
170 °C.15 This initial temperature screen was followed by a
stoichiometry screen in the same equipment set, but without
triethylamine.16 Ammonium acetate (1, 5, and 10 equiv) was
examined at 170 °C and 5 equiv was selected as optimal for this
temperature and residence time (<2% 11; >95% products).
Once again the entire range of experiments was conducted in a
single working day17 using just 5 g of ketoamide 11.18
The cyclization and deprotection sequence was scaled up to
20 g then to 775 g19 using the general equipment setup shown
Table 2. PFR cyclization results for ketoamide (11) cyclization and deprotection
Scheme 6. Cyclization of ketoamide 1 with known processing impurities
985
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
with acetonitrile resulted in clean, off-white solids with very
good handling properties and impurity profile.
A closer examination of several batch scale-up results (Figure 3)
revealed that reaction performance was declining as reaction scale
Probing temperature while holding constant Vreactor/Qfeed at
90 min indicated that a temperature range of 140−150 °C was
optimal at this reaction time with 5 equiv of ammonium acetate
(Figure 4). Shorter calculated residence times suffered from
incomplete conversion, and at higher temperatures extensive
loss of the protecting group was observed.
Preferred conditions were 10 equiv of ammonium acetate in
3 vol equiv of methanol at 140 °C with Vreactor/Qfeed = 90 min.
Once optimal conditions were determined, a series of scale-up
experiments (Table 3, entries 1−3) were conducted ahead of a
Table 3. Summary of cyclization of 1 using different plug
flow reactor sizes
AcONH (10 equiv), MeOH (3 vol equiv)
4
1 ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯→ 2
PFR, 140° C, V
Q
=90 min
feed
reactor
other
impurities
(%)
reactor size
(mL)
De-Boc
16 (%)
aniline
17 (%)
entry
2 (%) 1 (%)
1
2
3
4
4.51
143
79.13
80.34
81.75
82.9
4.4
4.32
4.73
3.35
5.7
2.02
7.02
6.6
8.14
5.03
3.82
4.39
0.51
1.32
1.03
1462
7140
4.56
GMP run (entry 4) where 55 kg of ketoamide 1 was processed
over 141 h at steady state in a 7.1 L PFR. The data in Table 3
represent average values at steady state. The consistency of
reaction performance across this wide range of reactor sizes can
be attributed to the fact that the reaction solution was a single
homogeneous phase, thermal expansion was the same at all
scales, axial dispersion was negligible for the coiled tube
reactors at all scales from 4.51 to 7140 mL (Table 10), and the
highest practical A/V was used at each scale to achieve
essentially isothermal conditions.
While the product profile average from the GMP campaign
looked very similar to the runs in smaller reactors (Table 3), close
review of the profile as a function of time indicated a subtle, but
steady linear decline in product area % (Figure 5). Likewise the
dimer impurity (18) trended higher over time. These two observa-
tions in combination with the fact that the starting material feed
solutions were prepared in advance and allowed to stand for the
prolonged time of the 6-day continuous run suggested that the
Figure 3. Trends in the batch cyclization of ketoamide 1.
increased. For example, area % of product 2 at 250-g scale was 10%
lower than at the initial 1-g scale reaction. This drop in product area
% was mirrored by an increase in other impurities as well. Since
heat-up times lengthen in batch reactors as scale increases, longer
time periods in an unproductive thermal regime would be antici-
pated during scale-up. It appeared as though lower-temperature
regimes (room temperature → 70 °C) favored greater degradation,
while at higher temperatures (>70 °C) cyclization was favored.
With a 25 kg delivery of 2 looming, we determined that traditional
batch scale up was not a viable option and we again chose to develop
the cyclization of 1 in a PFR. We were confident that this would
address any issues associated with heating rates and would also afford
precise control of time at reaction temperature. We also suspected
that we could improve the isolation of the product (2) by running in
methanol, a solvent amenable to direct crystallization.
Figure 4. Temperature screen for the cyclization of 1.22
986
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Figure 5. Product (2) area % trends in cyclization reaction in 7.1 L PFR.
ketoamide 1 was degrading over time in the presence of ammon-
ium acetate even at room temperature.
aromatic substitution under reaction conditions. A number of
different degradation pathways were possible with various
nitrogen nucleophiles as depicted in Scheme 7. However, it was
also clear that the imidazole products were not susceptible to
similar degradation pathways.23 This is not surprising since the
aryl fluoride would be much less activated toward SNAr with the
imidazole present and not the ketone.
The fact that long Vreactor/Qfeed times at 140 °C were optimal
for this reaction was interesting, given the rapid degradation
observed in Figure 6 . One possible explanation for why conver-
sion to product was favored at higher temperature was that
activation energy for conversion to product may be higher than
activation energies of the side reaction indicated by rate k2 or k3
(Scheme 8). The temperature dependence of a reaction rate
constant can be quantified by the Arrhenius Law.24 The larger the
activation energy for a reaction, the more temperature sensitive the
rate of reaction. This could cause the relative rates of conversion to
product and decomposition of 1 to be more favorable toward the
product 2 at higher temperatures.
We had conducted a brief stability study of 1 prior to the cam-
paign but not of a duration close to the time needed to complete
the campaign (141 h). In this study no noticeable degradation was
1
observed by HPLC area % or H NMR. Closer examination of
ketoamide 1 stability by wt % indicated a startling degradation
profile (Figure 6). What was even more surprising was that even at
Two sections of product (2) solution were produced from
the 7.1-L PFR reaction under GMP standards, and each section
was transferred to a pilot-scale batch reactor for workup and isola-
tion.25 The extraction, concentration, and addition of acetonitrile
and water afforded a technical grade product with about 80%
potency. The combined sections of crude imidazole product 2 were
then reslurried in acetonitrile to afford 29.2 kg of purified imidazole
in 52% isolated yield (potency corrected). Imidazole 2 was then
regioselectively alkylated with methyl iodide using potassium
hydroxide in DMSO to afford 3 in 90% isolated yield after crys-
tallization (Scheme 9). The crude product was recrystallized from
ethanol and water in 93% yield to remove aniline impurity 25
(Scheme 11). A total of 28.5 kg of 2 was processed through this
sequence to give 3 in 82% isolated yield. Methyl iodide was
controlled to <5 ppm in the product after recrystallization. Protect-
ing group removal (3→4) was accomplished in methanol using
anhydrous HCl (4 equiv) generated in situ from methanol and
acetyl chloride at room temperature. Over 99% conversion was
obtained in 13 h with no new observable impurities forming.
Solvent exchange from methanol to DMSO followed by treatment
Figure 6. Stability of 1 in the presence of ammonium acetate.
room temperature the steady drop in ketoamide 1 weight % was
observed with just a small change in HPLC area %. No significant
product formation or deprotection was observed at room tempera-
ture to account for the loss in potency. The solutions at elevated
temperature showed visible signs of degradation such as a dark-
orange/red color. Clearly, separate feed streams of ketoamide and
ammonium acetate would have been more appropriate. Subsequently,
it is standard practice to test stability of all reagent feed solutions for
the time durations that would be expected in a manufacturing facility.
On the basis of this stability study, it was clear that oligomeri-
zation pathways were operable. The observation of dimer 18 in-
dicated that the aromatic ring was activated towards nucleophilic
987
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Scheme 7. Possible SNAr-based degradation pathways
Scheme 8. Rationale for temperature-dependent degradation
Scheme 9. Synthesis of 5 from imidazole 2
with triethylamine afforded free amine 3, which was forward
processed directly in high yield in chemistry that will not be
discussed in this account.
While the overall sequence of this process was effective in
delivering material at kilogram scale in serviceable yield, there
were several undesirable aspects including the following:
(5) The synthesis of ketoamide 1 involved the use of bromide
7b (a strong lachrymator), azide (safety hazard), and a
Staudinger reaction which generates triphenylphosphine
oxide and nitrogen during the reaction.
Second Generation Approach using N-Methyl Keto-
amide 6. To address aspects 1 and 2 we sought a simple
structural change from ketoamide 1 to N-methyl ketoamide 6.
On the basis of the degradation pathways suggested in
Scheme 7 blocking the amide nitrogen would eliminate the
primary degradation pathway and allow for a much higher yield
of product. Regarding point 4, we had already observed thermal
cleavage of the Boc group under the cyclization conditions at
high temperatures. We wanted to further exploit the benefits of
the PFR and develop an efficient thermal protecting group removal
(1) The cyclization yield was low and required reprocessing
to afford material of acceptable quality.
(2) The use of methyl iodide, a compound that needed to be
controlled at ppm levels in 3.
(3) Aniline byproduct 18 must be rejected to low (ppm)
levels in intermediate 3.
(4) Removal of the Boc protecting group and salt break, while
high yielding, were inefficient and wasteful processes. The
use of 4 equiv of HCl was also not ideal.
988
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Scheme 10. Synthesis of N-methyl ketoamide 6 from N-Boc sarcosine (20)
Scheme 11. Cyclization of ketoamide 6
small-scale reaction afforded 80% of desired product 3 with aniline
levels dropping by about half to 0.35%. When acetic acid
(20 equiv) was added to a similar cyclization reaction in ethanol at
80 °C with 10 equiv of ammonium acetate, just 0.2% aniline could
be detected by HPLC after 18 h. The product (3) was obtained in
91% yield, and aniline levels were determined to be 29 ppm in the
isolated solid after crystallization. Overall, use of the N-methyl
substrate 6 (over ketoamide 1) cut aniline formation by about
10× (6% → 0.6%) and adding acetic acid to the cyclization of 6
dropped aniline formation another 3× (0.6% → 0.2%).
Given the issues with ketoamide 1 in the presence of ammo-
nium acetate at room temperature, a stability study was performed
with ketoamide 6. We were pleased to find that N-methyl ketoa-
mide 6 began to cyclize at room temperature in the presence of
ammonium acetate. Additionally, formation of cyclized product 3
tracked almost identically with wt % drop in ketoamide 6. This
supported the hypothesis that SNAr of the amide nitrogen was the
primary degradation pathway (Scheme 7).
While the new batch cyclization conditions in ethanol with
acetic acid were impressive this reaction suffered from two pro-
blems. First, long reaction times (16−20 h) were required to reach
completion. Second, much like previous examples, the best results
were those conducted in sealed tubes to prevent loss of ammonia.
In addition, ammonium salts are known to sublime which
may result in resolidification on colder surfaces in a batch tank.
The possibility of process piping in the reactor head (including
the reactor safety pressure relief) becoming blocked by solid
ammonium acetate seemed real.28,29 Addressing these two
problems was the initial goal of developing a continuous flow
process for this cyclization.
Initial cyclization PFR screening was conducted in a mixed
ethanol/methanol solvent system. Methanol was added to
improve solubility of ammonium acetate.30 Two feed solutions
were used: the first contained 6 in ethanol (3 mL/g), the
second contained ammonium acetate (10 equiv) and acetic acid
(20 equiv) in a mixture of EtOH (3 mL/g of 6) and MeOH
(1.5 mL/g of 6). With stable feed solutions in hand, we chose
Vreactor/Qfeed = 15 min as a starting point and temperatures from
100 to 190 °C were screened in 10 °C intervals. At tempera-
tures up to 130 °C, some starting material remained, but the
reaction was otherwise clean by HPLC. However, the cyclized
to eliminate any reagents in the reaction. Finally, we sought a rede-
signed route to N-methyl ketoamide 6 to eliminate the undesirable
reagents and unit operations in the current ketoamide synthetic
route (Table 1).
The synthetic route to N-methyl ketoamide 6 (Scheme 10)
was designed from N-Boc-protected sarcosine (20), a readily
available starting material that has the required methyl group
“pre-installed” avoiding any type of downstream alkylation. N-
Boc sarcosine (20) was converted to the morpholine amide 21
using standard conditions in modest yield. The crude amide 20
was treated directly with Grignard reagent (22)26 to afford
aminoketone 23. Treatment of 23 with HCl(g) in ethyl acetate
resulted in the salt of N-methylphenacylamine (24). The syn-
thesis was completed through an amide coupling between 24 and
10 with T3P in THF in 69% isolated yield. This unoptimized
process was used to synthesize 2 kg of N-methyl ketoamide 6
which funded scale-up studies in flow.
Development of a Second Generation Cyclization
Reaction from Ketoamide 6. Prior to PFR work on
ketoamide 6, it was found that adding acetic acid27 (10 equiv
with respect to starting ketoamide) to the cyclization conditions
reduced the formation of aniline 25 (Scheme 11). Ketoamide 6
also cyclized to 3 at a much faster rate than ketoamide 1 cyclized
to 2. The faster rate of cyclization with N-methyl ketoamide 6 can
be attributed to steric interactions which place the amide in a more
favorable conformation for cyclization. The faster rate of cycliza-
tion also affects the rate of aniline analogue 25 formation. For
example, treatment of ketoamide 6 with ammonium acetate
(10 equiv) in methanol (sealed tube) at 104 °C resulted in full
conversion in just 4 h. The amount of aniline analog 25 produced
during the reaction was 0.61% area by HPLC. The isolation of this
989
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Figure 7. Effect of residence time on yield (wt%) of 3 at 190 °C: thermal deprotection.
product 3 underwent thermal deprotection at a rate that increased
linearly as residence time increased at 190 °C (Figure 7).
Since the next step in the process was deprotection, we
considered the possibility of performing both cyclization and
deprotection reactions in the same PFR as a single telescoped
process. Unfortunately, at temperatures above 170 °C a new,
acylated impurity (26) began to form (Figure 8). In addition,
Figure 9. Cyclization in situ yield of 3 with respect to Vreactor/Qfeed and
temperature.
Isolation of crystalline 3 was accomplished by addition of
water as antisolvent to the reaction mixture. Later work showed
that comparable results were obtained with methanol as the
sole solvent, which simplified the reaction. The reaction rate
was slightly faster in methanol and the product could still be
crystallized directly by addition of water. The aniline impurity
(25) level in the isolated product from direct crystallization
(unoptimized) from methanol and water was <50 ppm. These
results were comparable to the batch process which provided 3
with 29 ppm of 25 under optimized conditions.
This process was successfully scaled up in a 221 mL PFR
with 1232 g of N-methyl ketoamide 6. The reaction was
conducted at 140 °C with a τ = 10.7 min. This provided 3 in
80% isolated yield (88% in situ). A comparison of the batch and
flow processes for the cyclization is summarized in Table 4.
Development of a Thermal Boc Group Deprotection
Reaction. The next task was to develop a PFR process that
could take advantage of the thermal Boc deprotection observed
previously in the cyclization reaction screening (Figure 9).
The N-Boc-protected imidazole 3 was dissolved in MeOH
Figure 8. HPLC traces for cyclization temperature screen at Vreactor
Qfeed = 15 min: 100 to 190 °C.
/
crystallization of 3 at this stage was the most efficient control
point for aniline impurity 25. For these reasons the cyclization
and thermal deprotection were developed as separate chemical
steps.
A residence time and temperature study for the cyclization of
6 was conducted at Vreactor/Qfeed ranges from 1 to 60 min and a
temperature range from 100 to 190 °C (Figure 9). In general,
temperatures <110 °C suffered from low in situ yields due to
incomplete conversion. Higher temperatures in the cyclization
led to more deprotection but when the amounts of product,
starting keto-amide and deprotected product were summed,
nearly all of the mass could be accounted for, indicating that the
polymerization was minimal. Optimal conditions of 140 °C and
(7 mL/g), and a temperature screen was conducted at Vreactor
/
V
reactor/Qfeed = 15 min were selected which maximized con-
Qfeed = 30 min. The desired thermal deprotection proceeded
but a new, unidentified impurity arose at temperatures >180 °C
(Figure 10). While the new impurity was not isolated and
identified, it seemed possible that it was the result of reaction
of 3 with MeOH since reactions in THF led to extremely
version while minimizing thermal deprotection.31 Under these
conditions the reaction was 80 times faster than the batch
reaction conditions and could potentially deliver production-
scale quantities in small PFR.32
990
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Table 4. Comparison of batch and flow processes for N−H cyclization
process
solvent
volumes
temp (°C)
time
isolation
25 (ppm)
yield (%)
batch
PFR
EtOH/HOAc
MeOH/HOAc
6
80
13−20 h
10 min
water addition
water addition
30
48
85
8033
6.5
140
Table 5. Impact of temperature on τ, D/uL of substrate 5,
Re, and density for flow-through 4.51 and 221 mL tube
reactors
a
reactor V (mL) temp, °C
τ
D/(uL)
Re
24
density (g/mL)
4.51
4.51
4.51
221
221
221
25
210
270
25
13.93
10.40
8.27
0.0011
0.0005
0.0006
0.0026
0.0004
0.0004
0.90
0.67
0.53
0.90
0.69
0.53
121
ND
15.93
12.21
9.35
293
210
270
1480
ND
a
Density is the average density for material inside the PFR during
steady-state flow conditions, calculated from measured F-curve,
numerical solution to calculate τ from F-curve, known V, and
measured mass flow rates.
Figure 10. Thermal deprotection in MeOH, HPLC profiles.
clean profiles. However, THF resulted in a slower reaction rate
than in methanol (∼5% conversion at 170 °C, Vreactor/Qfeed = 30
min) and the product (5) had low solubility in THF leading to
fouling at the reactor outlet. Using a 5% MeOH in THF solution
resolved the solubility problems, and this small amount of
methanol did not lead to measurable amounts of the unidentified
impurity previously observed in reactions using 100% MeOH.
The thermal deprotection process was run in a 4.51 mL PFR
and then in a 221 mL PFR to produce 1 kg of 5. Temperature
measurements at several points along the length of the 221 mL
tube verified that a temperature of >265 °C was reached within
<2% of the total reactor length as measured from the oven inlet.
The thermal deprotection reaction ran under supercritical fluid
conditions at 270 °C at 1000 psig with a measured τ = 9.4 min
in the 221 mL PFR.34 Concentration of the solution to a solid
afforded a quantitative yield of 5 of 97% potency. This material
was also crystallized from toluene and heptane to afford 79%
yield (unoptimized) of 5 with 99.5% potency. In practice,
however, the solution of 5 would be used directly in the
subsequent step in the chemical process.
Interestingly, the calculated mean residence time for the
thermal deprotection was Vreactor/Qfeed = 15.9 min for this
reaction, but the measured τ was just 12.2 min at 210 °C and
9.4 min at 270 °C (40% less than calculated if not correcting for
density changes in the reactor). As stated earlier in this paper
and remarkably clear in this case, when fluid density and/or
physical phase changes occur within the reactor compared to
feed solutions, Vreactor/Qfeed will not equal τ. In this specific case
the critical temperature for THF/MeOH was exceeded at
270 °C, resulting in a single-phase fluid with much lower
density (0.53 g/mL) vs. that at ambient (0.896 g/mL). At
210 °C which is below the critical temperature, the thermal
expansion of the liquid phase results in τ = 12.2 min (Table 5).
Calculating D/uL from the experimental F-curves in both the
4.51 mL (Figure 11) and 221 mL35 PFRs revealed that elevated
temperatures also have a dramatic effect on axial dispersion.
The F-curve measurements were made by first flowing solvent
only through the tube reactor at the given conditions, then
switching from solvent only to a solution of dissolved 5 flowing
into the tube, and measuring concentration of 5 vs. time
flowing out the end (Figure 11). In the figure, C/C0 is the ratio
of 5 concentration measured in tube effluent, C, divided by
concentration of 5 in the solution flowing into the reactor after
Figure 11. Experimental F-curves and dispersion model fits for
transition from solvent-only to dissolved substrate 5 at different tem-
peratures in a 4.51 mL PFR.
the step change, Co. D/uL decreases from 0.0011 to 0.0005 in
the 4.51 mL reactor, and D/uL decreases from 0.0026 to
0.0004 in the 221 mL reactor, when temperature is increased
from 20 to 210 °C. D/uL for supercritical conditions at 270 °C was
about the same as for 210 °C (Table 5). Definition of D/uL, how it
relates to axial dispersion, and calculation of D/uL from experi-
mental F-curves are described in the next section of this paper.
PFR systems are ideally suited to examine the effects of tem-
perature, stoichiometry, concentration, and τ on reaction yield
and purity. Combining automated temperature control (GC
oven) with the ability to automatically sample and dilute the
product stream at the reactor outlet for HPLC assay36 led to a
powerful automated reaction-condition screening setup. In this
study we examine the thermal Boc group deprotection of 5 as a
5% solution in N-methylpyrrolidinone (NMP) at Vreactor/Qfeed
=
15 min. A GC oven was programmed to heat to a given
temperature, hold for a set time (30 min), then heat to a new
temperature and hold again. The hold time could be varied to
establish a specific duration at each new steady state. The hold
time at each temperature was twice the Vreactor/Qfeed to ensure
at least two samples at steady state for every temperature. The
automated sampling and diluting system at the outlet provided
vials ready for wt % HPLC assay. Once started, the system
could run unattended for a specified duration.
991
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
A single-feed pump at a single flow rate was used in the first
dispersion which is well characterized in laminar flow tubes.38
The extent of dispersion affects the residence time that some
molecules are subjected to reaction conditions which can
influence impurity profile. Axial dispersion must also be under-
stood as the process is scaled to larger reactors that may have
different geometries or operate in different flow regimes.
Levenspiel showed that the vessel dispersion number (D/uL)
can be used to quantify the extent of dispersion in a given vessel
under a set of operating conditions, where D is the Taylor
longitudinal dispersion coefficient that incorporates the effects
of both diffusion and convection, u is average flow velocity, and
L is vessel length. A perfect plug flow reactor would have a D/
uL = 0, while a perfect continuous stirred tank reactor would
have D/uL = ∞. Using the guidance of Levenspiel, D/uL =
0.025 represents an intermediate amount of dispersion, while
0.002 represents a low amount of dispersion.39
experiment with this system, and therefore one value of Vreactor
/
Qfeed could be examined per screen. The GC oven was pro-
grammed to ramp from 180 to 260 °C. Aliquots were drawn
automatically every 10 min and assayed following the experiment.
This sequence ran completely unattended overnight, and the data
were analyzed the following day.37 In addition to conversion data
(Figure 12) that one might have also used a commercial React-IR
As axial dispersion increases, so does the reaction time
required for a given fractional conversion. Fractional conversion
of reactant A, as it passes through a tubular reactor, is governed
by axial dispersion, advection, reaction rate, and reaction order,
as shown in the following dimensionless form of the material
balance equation for component A (eq 1) where XA = fractional
conversion of reagent A, CA0= initial concentration, z = fraction
of total reactor length (l/L), k = reaction rate constant, and n =
reaction order.
Figure 12. Thermal De-Boc in NMP: automated temperature screening.
system to obtain, the HPLC samples allowed trace impurities to
be tracked (Figure 13), something not possible with React-IR. A
d2XA
dz2
dXA
dz
D
uL
−
+ kτCAn−1(1 − XA)n = 0
0
(1)
The impact of dispersion can be clearly seen in Table 6
where τ to reach 99.9% conversion for an isothermal elementary
Table 6. Impact of axial dispersion on reaction time for an
elementary first-order reaction that requires 90 min in an
ideal PFR
axial dispersion number
τ to reach 99.9% conversion change vs ideal
(D/uL)
(min)
(%)
0 (ideal)
90
0
0.002 (low dispersion)
91.2
105
+1.3
+17
Figure 13. HPLC traces from automated thermal De-Boc temperature
screen.
0.025 (intermediate
dispersion)
0.2 (higher dispersion)
196
+117
second generation system in which multiple pump flow rates as
well as temperatures can be automated has since been developed
and increases optimization speed in this reactor setup.
first order reaction has been calculated at multiple dispersion
numbers for a 90 min reaction. If axial dispersion is low (D/uL =
0.002), then τ required for 99.9% conversion is about 1.3% longer
than ideal PFR, assuming the same reaction conditions. Time
to reach 99.9% conversion would be 17% longer with inter-
mediate dispersion (D/uL = 0.025) and 117% longer with
higher dispersion (D/uL = 0.2). Low vessel dispersion numbers
ensure that the kinetic results obtained would be well modeled
by assuming plug flow reactor models (PFR) and thus neglect-
ing the first term in eq 1. Many reactions that chemists may
chose to run in flow are fast (seconds), such as metalation,
while others may be accelerated to achieve complete conversion
in seconds or minutes by extreme conditions. However, many
reactions require longer reaction times (hours) due to slower
kinetics or impurity formation under more extreme reaction
conditions. In the case of slower reactions, maintaining low axial
dispersion is particularly important to minimize the transition
time to steady state and minimize τ required for full conversion
at temperature that results in highest purity. The extent of axial
dispersion also influences impurity profile since some material
Safety of PFRs. One of the well-established benefits of
using a PFR is their inherent safety vs batch reactors. This
safety advantage arises from two primary areas, lower reactor
volume and superior heat transfer. Reactor volume in a PFR is
at least 1 order of magnitude smaller than in a batch reactor for
the same throughput, assuming one batch per day vs a
continuous reaction running with τ less than 2 h. In addition,
heat transfer surface area per unit volume (A/V) is 1−2 orders
of magnitude larger. Table 10 shows A/V calculations for each
of the PFRs used in this study, which can be compared to A/V
for typical batch reactors. The PFR setups used in this paper
were all tested at 1.5× the planned operating pressure. The
equipment setups were also equipped with pressure reliefs for
all positive displacement pumps and for the reactors. For highly
exothermic reactions, heat transfer fluid with turbulent mixing
or flow was used on the outside of the tube reactor, rather than
forced convection for PFR temperature control.
Continuous Reactor Selection and Characterization.
In a PFR, residence time distribution can be described by axial
992
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
resides in the reactor a longer time and some a shorter amount of
time. Maintaining negligible axial dispersion (plug flow) minimizes
this effect and can improve impurity profile in certain cases.
Plotting C/C0 vs t/τ after a step change in nonreactive tracer
flowing into a tube produces an S-shaped curve (for example,
see Figure 11), where C is nonreactive tracer concentration at
exit and C0 is tracer concentration at the inlet. Dankwerts
defined this as the F-curve.40 It is mathematically equal to the
integral of the Dankwerts “C-curve”, which is the C/C0
response curve to a pulse injection of nonreactive tracer. The
C-curve and the F-curve are functions of D/uL mathematically;
therefore, both may be used to calculate D/uL. The spread of
the C-curve as measured by its variance (σ2θ) is related to D/uL
by the relationship shown in eq 2:41
paper were coiled in various geometries, and the geometric factor
is certainly more complex than simply d/L. For example, Table 7
Table 7. Impact of coil diameter on axial dispersion
total tube length
(m)
d
coil diameter
(m)
D/uL (at τ = 90
entry
(mm)
min)
1
2
122
122
4.57
4.57
0.22
0.43
0.0010
0.0016
shows that axial dispersion number is slightly lower if the
tubing has a smaller coil diameter. Regardless, the trend
remains valid that coiled tubes with higher L/d have lower axial
dispersion, for about the same reactor V and τ (Table 8). For
D
uL
1
8
Table 8. Impact of L/d on axial dispersion
=
( 8σ2θ + 1 ) − 1
(2)
average V
(mL)
D/uL (at τ =
90 min)
D/uL (at τ =
V (mL)
L/d
5 min)
The relationship shown in eq 2 gives a convenient method
for quantifying D/uL from experimental F-curve data. In the
case of low dispersion the observed concentration change will
be normally distributed about a mean. Alternatively, D/uL can
be calculated by fitting experimental F-curve data to the basic
differential equation representing dispersion by numerically
solving for D/uL as shown in eq 3 where θ = dimensionless
time (t/τ).
a
1.8
25
1.75
12,773
435
0.0003
0.0065
0.0005
0.0010
0.0013
0.0003
0.0013
0.0047
0.0002
0.0016
0.0091
0.0008
0.0385
0.0017
0.0088
0.0004
0.0023
0.0098
0.0005
0.0039
0.0049
0.0001
0.0025
0.0031
0.0002
0.0019
1.84
a
20.5
149,587
9,822
28.8
25.5
682
776
632
1,714
700
1900
8000
105,813
10,336
3,106
2
⎛
⎜
⎞
⎟
a
∂C
∂θ
D
∂ C
∂z
∂C
∂z
1462
226,990
26,684
3,194
=
−
2
⎝
⎠
(3)
uL
2003
2185
D/uL was quantified by both methods described above for all
coiled tube reactors in this study. The two methods of
calculating D/uL from experimental F-curve data agreed within
5% where D/uL < 0.01. Where D/uL > 0.01, the reported value
was calculated by the variance method. All reactors were tested
using a 90 min τ as well as 5 min τ to span a typical range of
residence times for reactions reported in this paper. For each
solvent-only axial dispersion experiment, the F-curve was
obtained by making a step change from 100% THF to a
60%/40% (v/v) THF/toluene mixture flowing into the tube,
or vice versa, and monitoring the concentration of toluene in
the reactor effluent, either by online Raman probe or manual
sampling and off-line HPLC analysis. Axial dispersion was
measured at both τ = 5 and 90 min for 20 coiled tube reactors,
9 of which were used in this paper (Table 10). The average
axial dispersion number for the reactors used in this study was
0.0008, as characterized by the testing with solvents only. Using
the mathematical relationships introduced above, the difference
between real reactors with D/uL = 0.0008 vs ideal PFRs is
about 0.5% longer τ to reach the same fractional conversion.
On the basis of this data we refer to the tube reactors in this
study as PFRs for practical purposes, rather than PFDRs
(plug flow with dispersion reactors).
a
7140
19,553
2,845
8726
a
Coiled tube reactor used for running chemistries reported in this paper.
more details on these coiled tube reactors, see Table 10 for
complete characterization.
The impact of L/d is more dramatic for larger PFRs at
longer τ. These results can be explained with existing textbook
correlations,42 because D/uL is the product of d/L and D/ud,
and D/ud is a function of Re and Sc numbers with existing
correlations. For a given D/ud, Re and Sc, one way to decrease
D/uL is to decrease d/L. We opted for a conservative approach,
designing the reactors with high L/d in research and
development experiments when τ required to maximize
conversion and minimize impurities is initially unknown.43
Interestingly, there was less axial dispersion at slower flow
rates for the smaller tubes in the laminar flow regime, but there
was more axial dispersion at slower flow rates for the larger tubes.
F-curves for the smallest and largest reactor used for chemistries
reported in this paper are respectively shown in Figures 14 and
15. The figures show that the impact of τ on D/uL reverses when
we go from a 1.75 to a 7140 mL coiled tube.
The data in Table 10 supports that higher L/d leads to lower
axial dispersion over the wide range of reactor volumes and resi-
dence times shown in the table. Correlations for vessel dispersion
numbers have been developed of the general form shown in eq 4.42
Higher L/d also leads to faster characteristic mixing time by
diffusion, which is more important for the small-diameter
laminar flow tubes used in this study. For streamline laminar
flow in tubes, radial mixing is adequate compared to longi-
tudinal mixing to ensure a sufficiently uniform cross-sectional
concentration of material if the following relationship holds (eq 5,
Taylor criterion for sufficient radial mixing by diffusion alone)
is attained where Dv is the molecular diffusion coefficient.
D
uL
= (intensity of dispersion)(geometric factor)
(4)
The intensity of dispersion is defined as D/ud which has been
correlated to the Reynolds number (Re) and Schmidt number
(Sc), and for straight tubing the geometric factor is defined as d/L.
We more commonly refer to the inverse, L/d. The reactors in this
L
u
d2
28.8D
≫
(5)
v
993
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Higher L/d also leads to more isothermal reaction condi-
tions. The criterion for isothermal performance of a tubular reactor
in the laminar flow regime is given by the following constraint
on tube diameter (eq 6, Hickman criterion for isothermal
performance of tubular reactor).44 Thus, smaller inner diameter
leads to more isothermal characteristics for laminar flow tubular
reactors, as one would expect intuitively.
14kRgTw2
max(−ΔH)(20Ea − RgTw)
d2 ≤
R
(6)
where:
k=thermal conductivity of the reaction fluid (W m−1 K−1)
Rg=universal gas constant (8.314 J mol−1 K−1)
Tw=wall temperature (K)
Figure 14. Experimental F-curves and dispersion model fits for
1.75 mL tube reactor for τ = 5 and 90 min.
R
max=reaction rate maximum (mol m−3 s−1)
ΔH=heat of reaction (J mol−1)
Ea=activation energy (J mol−1)
One practical limit to higher L/d is pressure drop. Hagen−
Poiseuille calculations for laminar flow in tube reactors,45 as
well as measured pressure drop from the axial dispersion tests
at two flow rates are shown in Table 10. Average pressure drop
for solvents flowing in the tube reactors used in this study was
5 psi for τ = 90 min, and 100 psi for τ = 5 min. Most reactions
used by the pharmaceutical industry have longer reaction
times, which allow for high L/d PFRs with manageable pres-
sure drop.
CONCLUSIONS
■
In conclusion we have demonstrated two different synthetic
approaches to 1H-4-substituted imidazole 5. Both approaches
utilized PFRs for reaction optimization and scale up. The first
generation approach highlighted how continuous technologies
can be used to rapidly screen conditions and accelerate early
phase material delivery. Rapid heat up with PFR technology
allowed for the preparation of large-scale quantities of 2 in spite
of challenging starting material degradation kinetics. A pilot
plant-scale process in a 7.1 L PFR was described where 58 kg of
ketoamide 1 was processed through the key cyclization reaction
under cGMP conditions in a lab hood infrastructure. The
results in the PFR were superior to what could have been
obtained in batch equipment at this scale. The second genera-
tion effort highlighted an improved route to imidazole 5 via
N-methylketoamide 6. This approach highlighted a superior
thermal cyclization reaction and efficient and environmentally
friendly thermal Boc group deprotection reactions in PFRs.
Automated sampling and dilution for HPLC analysis
allowed for data-rich experimentation which led to faster
development and scale up. The cyclization of N-methylketoamide
6 and thermal deprotection reaction were both demon-
strated in PFRs at 1 kg scale. The deprotection was
accomplished under supercritical fluid conditions in the flow
reactor, which allowed for a single homogeneous phase
even though CO2 and isobutylene were evolved. The through-
put advantages in this chemistry also point to the possibility of
achieving low volume (1−2 MT) commercial scale throughput
safely in a laboratory hood infrastructure. PFRs represent
excellent cost-effective research tools for screening temperature,
extreme conditions, time, stoichiometry, but not necessarily
reagents or solvents. Overall, nine different PFR reactor sizes
were used in the course of this work. The high surface area per
unit volume, high pressure rating of tubes (vs tanks), and the
smaller size required for a given kg/day throughput make the
Figure 15. Experimental F-curves and dispersion model fits for 7140 mL
tube reactor for τ = 5 and 90 min.
Ideal plug flow behavior is defined by complete mixing in the
radial direction but no mixing in the axial direction. Therefore,
velocity profiles in the axial direction must be overcome by
radial mixing to approach plug flow behavior. Mixing in the
radial direction at large scale is dominated by turbulence as
represented by Re, but mixing in the radial direction at
smallest scale in laminar flow tube reactors is dominated by
diffusion. In the smaller-diameter tubes in the laminar flow
regime, axial dispersion is lower at longer τ, despite the fact
that Re is lower (Table 9). This is observed since longer τ
Table 9. Impact of Reynolds number on axial dispersion
Re (at
d
τ = 90
Re (at
D/uL (at
D/uL (at
V (mL)
(mm)
min)
τ = 5 min) τ = 90 min) τ = 5 min)
1.75
221
0.56
1.96
4.57
9.55
7.75
15.7
1.2
45
22
809
0.0003
0.0005
0.0013
0.0091
0.0008
0.0385
0.0017
0.0024
0.0039
0.0031
0.0002
0.0019
776
67
1215
1637
6598
3967
2185
7140
8726
91
367
220
gives more time for radial diffusion and compensates for
parabolic velocity profiles in the longitudinal direction.42 In
contrast, in the larger-diameter tube reactors, where diffusion
path in the radial direction is much longer and thus diffusion
has less impact on radial mixing, axial dispersion is lower at
higher Re (Table 9).
994
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
Table 10. Dimensions and characterization of coiled tubular PFRs
a
Dimensions and Dispersion
entry
V (mL)
o.d. (mm)
d (mm)
L (m)
L/d
D/uL (at τ = 90 min)
D/uL (at τ = 5 min)
b
1
1.75
1.84
2.40
3.68
4.51
17.0
20.5
25.5
28.8
143
1.59
3.18
1.59
3.18
1.59
3.18
1.59
3.18
3.18
3.18
3.18
3.18
3.18
6.35
9.53
3.18
6.35
6.35
12.70
9.53
19.05
0.56
1.75
0.56
1.75
0.56
2.67
0.56
2.67
1.55
1.96
1.96
2.02
2.02
4.57
6.38
2.02
4.57
4.57
9.55
7.75
15.7
7.14
0.76
9.79
1.52
18.4
3.05
83.6
4.57
15.2
47.6
73.6
169
12,773
435
0.0003
0.0065
0.0002
0.0035
0.0004
0.0020
0.0002
0.0013
0.0010
0.0010
0.0005
ND
0.0017
0.0088
0.0015
0.0063
0.0010
0.0081
0.0004
0.0098
0.0023
0.0024
0.0024
ND
2
b
3
17,517
870
4
b
5
32,918
1,143
6
b
7
149,587
1,714
8
9
9,822
b
10
24,344
37,622
83,974
105,813
10,336
3,106
b
11
221
b
12
541
13
14
15
682
213
0.0003
0.0013
0.0047
0.0002
0.0010
0.0016
0.0091
0.0008
0.0385
0.0005
0.0039
0.0049
776
47.3
19.8
458
632
b
d
16
1462
2003
2003
2185
7140
8726
226,990
26,684
26,684
3,194
0.0001
17
18
19
122
ND
122
0.0025
0.0031
0.0002
0.0019
30.5
151
b
20
19,553
21
44.8
2,845
c
Characterization
Re
Re
measured
calculated
pressure drop
(psi, at τ = 5 min)
measured
calculated
A/V
(at τ = 90 (at τ = 5
pressure drop
pressure drop
pressure drop
(m2/ number
entry
min)
min)
(psi, at τ = 5 min)
(psi, at τ = 90 min)
(psi, at τ = 90 min)
m3)
of coils
b
1
1.2
0.4
1.7
0.8
3.2
2.5
15
22
0
0
1
0.002
2
0
0
0.07
0
7158
2282
7158
2282
7158
1500
7158
1500
2577
2045
2045
1983
1983
875
20
2.0
27
2
7.5
b
3
31
15
0
0
0.1
4
0
0.006
9
0
0
1.6
65
b
5
58
6
2
0.5
6
46
0
0.01
180
0.02
0.8
5
0
0
2.9
131
4
b
7
263
69
160
0
9
10
8
9
3.8
7.4
29
0
0.001
0.04
0.3
133
524
809
ND
2420
1215
710
5191
3136
3136
1637
6598
3967
0
0
55
b
10
4
1
43
b
11
45
10
ND
226
0
11
1
0.6
88
b
12
ND
134
67
ND
90
ND
4
ND
5
ND
396
67
13
14
15
0.9
0.08
415
6
0
0.05
0.004
23
40
0
0
627
12
b
d
16
288
174
174
91
497
25
0
1983
875
675
180
90
17
18
19
13
14
0
0.3
6
0
0.3
875
0.08
3
0
0.005
0.2
419
18
b
20
367
220
28
0
1
516
119
17
21
0.07
0
0.004
254
a
b
V is tube volume, o.d. is tube outer diameter, d is tube inner diameter, L is tube length, D/uL is vessel dispersion number. Coiled tube reactor used
c
for running chemistries reported in this paper. Re is Reynolds number for solvent tests at room temperature; A/V is heat transfer surface area per
unit volume. Planned τ was 5 min, but actual τ was 9.45 min.
d
PFRs intrinsically safer than batch processing for the thermal
reactions described herein. The ability to access extreme
conditions in these reactors expanded the design space
compared to conventional laboratory and scale-up equipment.
D/uL for the nine coiled tube reactors used in this study
averaged 0.0008 and ranged from 0.0001 to 0.0024, as
characterized by testing with solvents only at τ = 5 and
90 min. Therefore axial dispersion is negligible and the tubes
can be modeled as PFRs with errors less than 2% between the
actual and the theoretically ideal.
EXPERIMENTAL SECTION
■
Materials. Starting materials, solvents, and reagents were
purchased commercially and used without further purification
unless otherwise indicated. HPLC analysis performed using
1
Agilent 1100 HPLC system unless otherwise indicated. H
NMR, 19F, and 13C NMR spectra were obtained on a Varian
spectrometer at 400 and 100 MHz and calibrated using residual
undeuterated solvent as an internal reference. The following
abbreviations were used to explain the multiplicities: s = singlet,
995
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
d = doublet, t = triplet, q = quartet, quin = quintuplet, m =
multiplet, br = broad. IR spectra were obtained on a Smiths
ChemID FTIR and high-resolution MS data was acquired on
an Agilent G1969A MS-TOF. Melting points were taken on a
DSC Q1000 machine or obtained via DSC analysis. The
dimensions of the PFR reactors used in this paper are as follows
in Table 10. All reactors were constructed with 316 L SS.
THF (4 L). The solids were dried in a vacuum oven at 40 °C
overnight to afford 2340 g (72%) of (2-amino-1-(3-trifluo-
romethyl-phenyl)-ethanone)-, p-toluene sulfonate (1:1) as a
1
white, crystalline solid (mp 197 °C). H NMR (400 MHz,
DMSO-d6) δ 8.24−8.30 (m, 5H), 8.09 (d, 1H, J = 8 Hz), 7.83
(t, 1H, J = 8 Hz), 7.46 (d, 2H, J = 6.4 Hz), 7.08 (d, 2H,
7.6 Hz), 4.69 (brs, 2H), 2.26 (s, 3H); 19F NMR (376.2 MHz,
DMSO-d6) δ −61.273 (s, 3F); IR (KBr, ν/cm−1) 3154, 2998,
2905, 1703, 1495, 1331, 1225, 1172, 1115, 812, 694, 555. Elem.
Anal.: Calcd for C16H16F3NO4S: C, 51.20%; H, 4.30%, N,
3.73%. Found: C, 51.24%; H, 4.39%, N, 4.00%.
2-Bromo-1-(3-trifluoromethyl-phenyl)ethanone46
(7a). To a 22 L four-necked round-bottom flask (fitted with
addition funnel, nitrogen blanket, condenser, scrubber, and
mechanical stirrer) was added 3-trifluoromethylacetophenone
(1500 g, 1.00 equiv, 7.97 mol) and dichloromethane (7.5 L).
The resulting clear, colorless solution was stirred at room
temperature while adding a solution of bromine (1274 g, 1.00
equiv, 7.97 mol) in dichloromethane at room temperature via
addition funnel over 4 h. After sampling to determine reaction
completion by GC47 the reaction was quenched by slow addi-
tion of saturated aqueous NaHCO3 (2000 mL), controlling the
temperature by ice bath to less than 25 °C. The phases were
separated, and the organic layer washed with brine (2000 mL);
the solution was then dried over sodium sulfate, filtered, and
concentrated on a rotovapor to a clear colorless oil. This crude
oil, containing the product plus 7−8% each of the starting
acetophenone and dibrominated byproduct, was purified by
silica gel chromatography (silica gel; step gradient, 20% to 50%
CH2Cl2 in heptane) to afford bromide 7a (1667 g, 6.24 mol,
4-[2-Oxo-2-(3-trifluoromethyl-phenyl)-ethylcarbamo-
yl]-piperidine-1-carboxylic Acid tert-Butyl Ester (11a).
Amine salt 15a (913 g, 2.43 mol) and 10 (623 g, 2.73 mol,
1.12 equiv) along with THF (2.75 L) and EtOAc (5.5 L) were
combined in a 22 L three-necked RBF fitted with water-cooled
condenser, nitrogen blanket, mechanical stirrer, and cooling
bath flask. The reaction was cooled to 0−5 °C with an ice bath.
T3P (1.2 equiv, 2.91 mol, 1.512 L of a 50% solution in EtOAc)
was added over 15 min and the reaction temperature main-
tained at 0−5 °C during addition. After stirring for 10 min N-
methylmorpholine (566 g, 5.6 mol, 2.3 equiv) was added over
30 min, holding the temperature below 5 °C during addition.
The reaction was warmed to room temperature, and once
deemed complete (<5% 15a by HPLC, usually after 10 h), the
mixture was cooled in an ice bath, and water (7.3 L) was added
rapidly. The layers were separated, and the aqueous layer was
washed with EtOAc (2.75 L). The organic layers were
combined and washed with 0.5 M aqueous sodium bicarbonate
solution (2.75 L). The organic layer was washed with brine
solution (2.75 L), and the resulting organic layer was then
treated with sodium sulfate and filtered. The solvent was
removed via atmospheric distillation down to ∼4.6 L. Heptane
(11 L) was added while removing solvent via distillation until
the final volume reached ∼11 L. The mixture was cooled to
50 °C and seeded. The resulting slurry was held at 50 °C for
3 h and then cooled to room temperature and stirred for 2 h.
The solids were filtered, and the cake was washed, first with
10% EtOAc in heptane (2 L) and then with heptane (2 L).
After drying in a vacuum oven at 60 °C for 3 h, 800 g (79%) of
1
78%) as a clear, colorless oil (100% purity by GC). H NMR
(400 MHz, CDCl3) δ 8.24 (s, 1H), 8.17 (d, 1H, J = 8.4 Hz),
7.87 (d, 1H, J = 8 Hz), 7.66 (t, 1H, J = 8 Hz), 4.46 (s, 2H)
(2-Amino-1-(3-trifluoromethyl-phenyl)-ethanone)-,
p-Toluene Sulfonate (15a). Bromide 7a (1664.7 g, 1.00
equiv, 6.23 mol) and tetrahydrofuran (7500 mL) were charged
to a 12 L three-necked RBF fitted with water-cooled condenser,
nitrogen blanket, mechanical stirrer, and cooling bath. Sodium
azide (425.6 g, 1.05 equiv, 6.55 mol) was charged in one por-
tion as a solid along with water (135 mL). The pale-yellow
slurry was stirred at room temperature under nitrogen. After
6 h, water (260 mL) was added, and the mixture was stirred
overnight. The resulting orange slurry was then filtered over a
thin pad of Celite and rinsed with THF (1 L). The resulting
solution of intermediate azide was divided into two equal por-
tions (5146.5 g each) for side-by-side reductions. Two identical
22 L three-necked RBFs (fitted with addition funnel, water-
cooled condenser, nitrogen blanket, mechanical stirrer, and
cooling bath) were charged with triphenylphosphine (889 g,
3.43 mol, 1.1 equiv), p-toluene sulfonic acid monohydrate
(1304 g, 6.86 mol, 2.2 equiv), and THF (5.6 L). The two
portions of intermediate azide mixture were added via addition
funnels to the individual flasks over 4 h, controlling the foaming
from nitrogen evolution by addition rate and the temperature
to <20 °C by use of an ice bath. CAUTION: Vigourous gas
evolution occurred during this addition. On completion of addi-
tion the slurry was stirred at ambient temperature for 2 h, and
then the solids were filtered from both reactors onto the same
filter. Both flasks and the combined cake were washed with
1
ketoamide 11a was obtained as a white solid (mp 116 °C). H
NMR (500 MHz, DMSO-d6) δ 8.30 (t, 1H, J = 5.5 Hz), 8.25
(d, 1H, J = 7.5 Hz), 8.21 (s, 1H), 8.03 (d, 1H, J = 7.5 Hz), 7.79
(t, 1H, J = 7.5 Hz), 4.61 (d, 2H, J = 5.5 Hz), 3.91 (d, 2H, J =
12.5 Hz),2.75 (brs, 2H), 2.40−2.51 (m, 1H), 1.66 (dd, 2H,
996
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
J = 13.5, 3 Hz), 1.30−1.44 (m, 9H); 19F NMR (376.2 MHz,
DMSO-d6) δ −61.342 (s, 3F); IR (KBr, ν/cm−1) 3229, 3011,
2942, 2851, 1642, 1544, 1429, 1368, 1160, 1066, 805, 690;
HRMS (ESI): calc’d for C20H25O4N2F323Na1 = 437.1659, found
437.1647; See Supporting Information for HPLC chromatogram.
(4-[4-(3-Trifluoromethyl-phenyl)-1H-imidazol-2-yl]-pi-
peridine)-, dihydrochloride (13a). A solution of ketoamide
11a (773.3 g, 1.00 equiv, 1.87 mol) in methanol (2300 mL)
was prepared. (Total solution volume = 2900 mL, [s.m.] =
1.87 mol/2.90 L = 0.645 M.) A separate solution of ammonium
acetate (755.08 g, 9.80 mol) in methanol (3500 mL) was
prepared. (Total solution volume = 4137 mL, {NH4OAc] =
9.80 mol/4.137 L = 2.367 M.) See Figure 1 for a representation
of the reactor setup. Ketoamide 11a solution was pumped at
5.088 mL/min and the NH4OAc solution at 6.932 mL/min,
resulting in a 5:1 mol ratio of NH4OAc to 11a. The two
streams were combined in a tee at room temperature and then
allowed to flow into the thermal tube reactor described below
with the oven temperature at 170 °C. The total time required
to process all of the starting material at these flow rates was
approximately 5 h.
reactor was added THF (350.0 kg) while maintaining a
temperature range of 15−25 °C. Bromide 7b (78.5 kg) was
then added to the reactor. The mixture was cooled to −2 to
2 °C, and an aqueous solution of sodium azide (prepared with
sodium azide (19.7 kg) and purified water (62.8 kg)) was
added to the mixture at the rate of 15−20 L/h. The reaction
was sampled every hour and analyzed by HPLC for completion
(<0.5% 7b). After stirring 10 h at −2−2 °C the reaction was
deemed complete by HPLC. The stirring was stopped, and the
layers were allowed to settle for 30 min before removing the
aqueous phase. The resulting crude solution of azide 14b was
used without isolation in the next unit operation.
In a separate 3000 L reactor were added THF (351 kg),
triphenylphosphine (86.7 kg), and p-toluene sulfonic acid
(104.8 kg) in sequence. The mixture was stirred for 30 min at
room temperature. The crude solution of azide 14b was added
to the reactor at a rate (30−40 L/h) to maintain an internal
temperature between 20 and 25 °C. The addition rate also
controlled the extent of nitrogen off-gassing. The product (15b)
crystallizes from solution during this time to form a slurry. After
the mixture was stirred 22.5 h azide intermediate 14b was still
present at >0.2%. Additional triphenylphosphine (3.6 kg in
9.6 kg THF) was added to the mixture; 3 h later, the reaction
contained <0.2% of azide 14b.
Then, maintaining the temperature at 20−25 °C, purified
water (7.9 kg) was added to quench the reaction at the rate of
10 kg/h. The slurry was stirred for 1 h and then was cooled to
0−5 °C and stirred for 3 h. The mixture was filtered by centrifuge,
and the cake was washed twice with THF (209.6 kg × 2) which
was cooled to 0−5 °C in advance. The filter cake was dried at
50−55 °C until the water content was <0.5% and LOD <0.5%.
15b (97 kg, 89.6%, 99.7% purity) was obtained as an off-white
solid. Amine salt 15b: White crystalline solid, mp 212−219 °C.
1H NMR (399.84 MHz, DMSO-d6) δ 8.40−8.36 (m, 1H), 8.32−
8.27 singlet-broad, 3H), 7.78−7.73 (m, 1H), 7.45 (d, J = 8.3 Hz,
2H), 7.08 (d, J = 7.5 Hz, 2H), 4.68 (s, 2H), 2.26 (s, 3H).
Desired Vreactor/Qfeed was 45 min in the 541 mL tube reactor,
so total flow rate = 12.02 mL/min was calculated. Reaction
profile was periodically monitored by HPLC throughout the
5 h run time. Typical steady-state reaction mixtures contained
<1% 11a, ∼25% 13a freebase resulting from thermal
deprotection and ∼75% 12.
All of the collected product solution was combined and
concentrated on rotovap and solvent exchanged into 3500 mL
of n-BuOH. This solution was washed with a mixture of
3000 mL of sat’d aqueous NaHCO3 and 1000 mL water.
This n-BuOH sol’n was then washed with sat’d aqueous NaCl
(3000 mL) followed by azeotropic drying by distillation of 1 L
of n-BuOH on the rotovapor (45 °C bath temp). The solids
(NaCl) that precipitated during distillation were filtered, and
the filtrate was charged to a 5 L four-neck flask in a cooling
bath. Anhydrous HCl (g) was bubbled slowly into the solution,
controlling the temperature <55 °C with and ice−water bath.
Solids precipitated during the HCl gas addition. Reaction
progress was monitored by HPLC until <0.5% of the Boc-
protected intermediate remained. On completion of reaction
heptane (4000 mL) was added slowly via addition funnel, and
the resulting slurry was cooled to <5 °C in an ice bath and held
for 15 min. The solids were filtered, and the cake was washed
with heptane (2 × 600 mL). Solids were dried in a 40 °C
vacuum oven to afford 619.1 g (92%) of imidazole salt 13a as a
19F NMR (376.2 MHz, DMSO-d6) δ −60.324 (d, J = 11.99 Hz,
3F), −107.17 (m, 1F); HRMS (ESI): calcd for C9H8F4NO =
222.0537, found: 222.0531; IR (KBr, ν/cm−1) 3436, 2913, 1669,
1621, 1503, 1327, 1254, 1172, 1037, 1008, 816, 666; Elem. Anal.:
calcd for C16H15F4NO4S: C, 48.86%; H, 3.84%, N, 3.56%. Found:
C, 48.89%; H, 3.88%, N, 3.63%
4-[2-Oxo-2-(4-fluoro, 3-trifluoromethyl-phenyl)-ethyl-
carbamoyl]-piperidine-1-carboxylic Acid tert-Butyl Ester
(11b). To a 1000 L glass-lined reactor was added ethyl acetate
(416.9 kg) while maintaining an internal temperature of 20−
25 °C. The mixture was then sampled to measure the water
content. An additional portion of THF (104.7 kg) was then
1
white solid; H NMR (400 MHz, DMSO-d6) δ 9.28−9.21 (s,
1H), 9.20−9.12 (s, 1H), 8.34 (s, 1H), 8.26−8.25 (m, 2H),
7.76−7.70 (m, 2H), 3.49−3.43 (m, 1H), 3.37 (q, J = 6.4 Hz,
2H), 3.08−2.99 (m, 2H), 2.30 (d, J = 12.3 Hz, 2H), 2.16−2.05
(m, 2H); IR (KBr, ν/cm−1) 2492, 2132, 1703, 1569, 1454,
1339, 1286, 1168, 1123, 956, 694, 547; HRMS (ESI): calcd
for C15H17N3F3 = 296.1369, found 296.1360; See Supporting
Information for HPLC chromatogram.
(2-Amino-1-(4-fluoro, 3-trifluoromethyl-phenyl)-etha-
none)-, p-Toluene Sulfonate (15b). To a 1000 L glass-lined
997
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
added. Carboxylic acid 10 (40.6 kg) was added to the mixture
under nitrogen gas, and the contents were stirred for 0.5−1 h
until all the solids dissolved. Stirring was stopped, and the
reactor was evacuated to <0.06 MPa and refilled with nitrogen
gas to reach ambient pressure (repeated three times). Stirring
was restarted, and 1,1′-carbonyldiimidazole (31.1 kg) was added
to the mixture under the protection of nitrogen gas at 20−25 °C.
The reaction was stirred for 30 min before 2627394 (58.0 kg)
was added to the mixture followed by N-methylmorpholine
(37.4 kg) at a rate of 6−8 kg/h. The reaction was stirred 16.5 h
at 20−25 °C. The reaction was considered complete when the
assay yield was >78% and the purity of 11b was >85%.
79.6 mL/min (Vreactor/Qfeed = 90 min). As the solution exited
the oven, it was cooled back to 20 °C in a tube-in-tube heat
exchanger.
Once the entire solution had been processed through the
reactor (72 h total processing time), the resulting orange
solution was split into two equal sections for workup as follows:
The solution was concentrated under vacuum (50−75
mmHg at 15−30 °C) to a total volume of 165 L. Acetonitrile
(60 L) was then added and the solution heated to 50 °C. Water
(185 L) was added slowly with seeding over 1 h to crystallize
the product. The resulting slurry was stirred at 50 °C for 1 h
and then cooled to 20 °C and stirred for 2 h. The solids were
filtered and washed with 20% MeOH in water (2 × 65 L). The
resulting solids were dried under vacuum at 50 °C.
Maintaining the temperature at 20−30 °C, water (290.0 kg)
was charged to a clean glass-lined reactor. Then the reaction
mixture was added to the water and stirred for 10 min. Stirring
was stopped, and the mixture was allowed to settle for 30 min
before the aqueous phase was removed. Ethyl acetate (52.2 kg)
was added to the aqueous phase and was stirred for 30 min.
The phases were allowed to separate over 30 min, and the
aqueous phase was removed. The combined organic phases
were washed with aqueous citric acid (23.3 kg of citric acid and
208.8 kg of water) for 30 min before allowing the phases to
separate (30 min). The organic phase was then washed with
brine solution (145.1 kg) for 30 min before allowing the layers
to separate (30 min). The organic layer was then concentrated
to 140−150 L under reduced pressure (0.085 MPa) while
maintaining the temperature below 35 °C. The mixture was
cooled to 30 °C over 2−3 h, then cooled to −3 to 3 °C for 7 h
for crystallization. The mixture was filtered by centrifuge. While
the temperature was maintained at 20−30 °C, the filter cake
was washed with n-heptane (78.9 kg) and dried at 40−45 °C
for 23 h. This afforded ketoamide 1 (46.7 kg; 73.2% yield) as an
off-white solid in 99.5% purity by HPLC. 11b: white solid, mp
141−144 °C; 1H NMR (399.84 MHz, DMSO-d6) δ 8.35−8.31
(m, 1H), 8.29−8.24 (m, 2H), 7.71−7.66 (t, 1H), 4.59 (d, J =
5.7 Hz, 2H), 3.89 (d, J = 12.7 Hz, 2H), 2.74−2.63 (m, 2H),
2.44−2.37 (m, 1H), 1.65 (dd, J = 2.6, 13.2 Hz, 2H), 1.37 (s,
9H); 19F NMR (376.2 MHz, DMSO-d6) δ −60.324 (d, J =
12.18 Hz, 3F), −108.936 (m, 1F); IR (KBr, ν/cm−1) 3146,
2946, 1675, 1618, 1503, 1417, 1368, 1323, 1279, 1242, 1050,
841, 666; Elem. Anal.: calcd for C20H24F4N2O4: C, 55.55%; H,
5.59%, N, 6.48%. Found: C, 55.50%; H, 5.46%, N, 6.50%;
tert-Butyl 4-(4-(4-fluoro-3-(trifluoromethyl)phenyl)-
1H-imidazol-2-yl)piperidine-1-carboxylate (2). A solution
was prepared by combining ketoamide 11b (55.175 kg, 127.7
mol), ammonium acetate (110.25 kg, 1430.33 mol, 11 equiv),
and methanol 580 L). The reactor used for this transformation
was a coiled stainless steel tube with 7.75 mm i.d. and 7.14 L
volume (Table 10) that was heated in an oven48 to 140 °C. The
back pressure in this tube was controlled at 640 psig by a
regulator to allow superheating of the solution above its normal
boiling point. The solution prepared above was pumped
continuously (using high pressure syringe pumps equipped
with automated valve switching packages to allow continuous
pumping and refill) through the heated tube under pressure at
The dried, crude solids from both sections above were then
combined and reslurried in acetonitrile (155 L) at 50 °C. The
slurry was cooled to ambient temperature, and the solids were
filtered and washed with acetonitrile (120 L) to afford
imidazole 2 (29.20 kg, 70.6 mol, 52.5% yield- potency
1
corrected) as an off-white solid (mp 199−202 °C). H NMR
(399.84 MHz, DMSO-d6) δ 11.96 (s, 1H), 8.03 (d, J = 6.6 Hz,
2H), 7.67 (s, 1H), 7.44 (t, J = 9.9 Hz, 1H), 3.97 (d, J = 13.2 Hz,
2H), 2.91−2.83 (m, 3H), 1.88 (dd, J = 2.4, 13.0 Hz, 2H), 1.62−
1.51 (m, 2H), 1.39 (s, 9H); 19F NMR (376.2 MHz, DMSO-d6)
δ −61.186 ppm: two imidazole tautomers observed in a 12:1
ratio: major tautomer δ −60.01 (d, J = 11.99 Hz, 3F), −120.642
(m, 1F); minor tautomer δ −59.96 (d, J = 12.00 Hz, 3F),
−119.575 (m, 1F); IR (KBr, ν/cm−1) 3199, 3109, 2953, 2917,
2856, 1654, 1540, 1442, 1344, 1131, 833, 763, 677; Elem. Anal.:
calcd for C20H23F4N3O2: C, 58.11%; H, 5.61%, N, 10.16%.
Found: C, 58.14%; H, 5.64%, N, 10.20%; HRMS (ESI): calcd
for C20H23F4N3O2 414.1799, found 414.1792.
tert-Butyl 4-(4-(4-fluoro-3-(trifluoromethyl)phenyl)-1-
methyl-1H-imidazol-2-yl)piperidine-1-carboxylate (3).
Compound 2 (35 kg, 84.66 mol) was dissolved in dimethyl
sulfoxide (455 L). KOH (85%, powdered) (7.99 kg, 121.0 mol,
1.42 equiv) was added in one portion. The result was a thin
slurry. Methyl iodide (12.6 kg, 88.77 mol, 1.05 equiv) was
charged over 30−60 min pump, maintaining temperature <30 °C.
The resulting solution was stirred at 25−30 °C for 1 h or until 2
was consumed to <3% area by HPLC.
A mixture of 3 seed crystals (0.07 kg) and water (105 L) was
added over 10−20 min to saturate the solution. The resulting
thin slurry was allowed to stir at 25 °C for 20−40 min, at which
time the slurry thickened. Additional water (315 L) was then
added over 60 min at 23−27 °C to complete the crystallization.
The solids were filtered and washed with 20% DMSO in water
(2 × 52.5 L and 1 × 105 L) and then dried under vacuum at
60 °C. The resulting dried solids were dissolved in ethanol
(245 L) at 50 °C. Water (97.7 L) was then added over 5 min.
Seed crystals of 3 (0.065 kg) were then added, and more water
(97.7 L) was added over 30−45 min to complete the
crystallization. The resulting slurry was cooled to 25 °C over
2 h. The solids were filtered, and the cake was washed with 20%
EtOH in water (2 × 83 L). The solids were dried under
998
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
vacuum at 60 °C, affording 31.10 kg (72.76 mol, 86%) of
methyl imidazole 3 as a white solid (mp = 199−202 °C). H
Na2SO4. The solution was filtered, and the filtrate was con-
centrated under reduced pressure at 50−55 °C to afford 21
(1370 g, 5.3 mol, 67% yield) as a yellow oil. The crude
mixture was used in the next step without further purification.
1H NMR (CHCl3) δ 4.03 (s, 2 H), 3.65 (s, 4 H), 3.59
(d, 2 H, J = 3.2 Hz), 3.43 (s, 2 H), 2.91 (s, 3 H), 1.43 (d, 9 H,
J = 14.0 Hz); 13C NMR (CHCl3) δ 80.1, 66.8, 66.5, 50.0, 45.2,
42.1, 35.4, 28.3; IR (CHCl3) 2973, 2857, 1696, 1663, 1456,
1177, 1115, 1032, 772 cm−1; Exact Mass: Calcd m/z for
1
NMR (399.84 MHz, CDCl3) δ 7.93−7.87 (m, 2H), 7.15 (d-d,
1H), 7.06 (s, 1H), 4.24−4.22 (m, 2H), 3.65 (s, 3H), 2.88−2.80
(m, 3H), 2.00−1.84 (m, 4H), 1.71−1.61 (m, 3H), 1.47 (s, 9H);
19F NMR (376.2 MHz, DMSO-d6) δ −61.37 (d, J = 12.57 Hz,
3F), −118.38 (m, 1F); IR (KBr, ν/cm−1) 3436, 2938, 1679,
1425, 1364, 1336, 1262, 1237, 1160, 1123, 914, 829, 690; Elem.
Anal.: calcd for C21H25F4N3O2: C, 59.01%; H, 5.90%, N, 9.83%.
Found: C, 59.06%; H, 5.88%, N, 9.83%.
+
C12H23N204 : 259.1652, Found: 259.1649.
4-(4-(4-Fluoro-3-(trifluoromethyl)phenyl)-1-methyl-
1H-imidazol-2-yl)piperidine (5). An anhydrous HCl sol-
ution was prepared by slow addition of acetyl chloride (16.2 kg,
206.4 mol, 4.00 equiv) to methanol (88 L) over 45 min at
<5 °C. The resulting solution was then added to a separate flask
containing a solution of 3 (22 kg, 51.5 mol) in methanol (88 L)
over 90 min at 20 °C, rinsing forward with MeOH (12 L). The
reaction mixture was stirred at 20 °C for 14 h at which point
HPLC analysis showed no 3 remaining. Dimethyl sulfoxide
(115 L) was then added, and the reaction mixture was then
concentrated under vacuum (150 mmHg at 30 °C) to a total
volume of 110 L. Dimethyl sulfoxide (29 L) was then added
back to the tank in one portion, and the distillation resumed.
DMSO (84 L) was then added continuously to the tank during
the distillation to maintain a constant volume of 139 L. The
pressure was maintained at 20 mmHg. The temperature of the
solution gradually increased from 5 to 6 °C at the beginning of
distillation to 50−55 °C by the end of distillation. Dimethyl
sulfoxide (125 L) was then added, and the solution was cooled
to 20 °C. At this point the reaction mixture should contain
<500 ppm MeOH. Triethylamine (21.7 kg, 214 mol, 4 equiv) is
then added over 45 min at 15−25 °C to yield 5 as a solution in
DMSO for use in subsequent chemistry. Compound 5 was also
crystallized and fully characterized. For these data see the
kilogram run in the PFR later in the Experimental Section.
tert-Butyl Methyl(2-morpholino-2-oxoethyl)carba-
mate (21). To a 22 L three-neck round-bottom flask (RBC)
equipped with N2 sweep was added 20 (1500 g, 7.93 mol) and
ethyl acetate (15 L). 1,1′-Carbonyldiimidazole (CDI) (1542 g,
9.52 mol) was added as a solid to the mixture over 10 min at
room temperature. The reaction mixture was stirred at room
temperature for 3−4 h. Morpholine (829 g, 9.52 mol) was then
added, and the reaction mixture was stirred at room
temperature for an additional 16 h. Water (7.5 L) was added
to the reaction mixture, and the layers were separated. The
aqueous layer was extracted with ethyl acetate (2 × 7.5 L). The
combined organic layers were washed with 1 N HCl (7.5 L),
followed by water (7.5 L), and then were dried over 700 g of
1-(4-Fluoro-3-(trifluoromethyl)phenyl)-2-(methylami-
no)-ethanone Hydrochloride (24). To a 2 L three-neck RBF
equipped with N2 sweep was charged 4-bromo-1-fluoro-2-
(trifluoromethyl)benzene (103.3 g, 425 mmol) and THF
(175 mL). The reaction mixture was purged with N2 twice.
Isopropyl magnesium chloride (2 M solution in THF, 203 mL,
406 mmol) was added dropwise over 20 min at room temp-
erature. The reaction mixture was then stirred at room
temperature for 2−3 h resulting in 22. A solution of 21 (50 g,
194 mmol) in THF (50 mL) was added dropwise over 10 min at
25−45 °C. The reaction mixture was stirred at room temperature
for 5 h. The reaction was then quenched with saturated NH4Cl
solution (50 mL). Ethyl acetate (600 mL) was added, followed by
water (250 mL). The pH of the aqueous layer was adjusted to
5−7 with 1 N HCl. Upon acidification, clear layer separation was
observed. The layers were separated, and the aqueous layer was ex-
tracted with ethyl acetate (200 mL). The combined organic layer
was washed with brine (250 mL) and dried with Na2SO4 (100 g).
The organic layer was then filtered, and the filtrate was concen-
trated under reduced pressure to afford 23 (70.9 g, 211.5 mmol,
999
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
109% recovery) as a dark-brown oil, which was used in the next
step without further purification. To a three-neck (RBF) equipped
with N2 sweep was charged 23 (70.9 g, 211.5 mmol) and ethyl
acetate (35 mL). The reaction mixture was cooled to
0−10 °C using an ice bath. HCl (2.0 M solution in ethyl acetate,
994.1 mmol) was added dropwise at a rate sufficient to maintain
the temperature at 0−10 °C. The reaction mixture was then
stirred under N2 at room temperature for 16−20 h. The reaction
mixture was filtered, and the cake was rinsed with 0−10 °C ethyl
1056 cm−1; Exact Mass: Calcd m/z for C21H26F4N2O4Na+:
469.1721, Found: 469.1712. See Supporting Information for
HPLC chromatogram.
tert-Butyl 4-(4-(4-Fluoro-3-(trifluoromethyl)phenyl)-1-
methyl-1H-imidazol-2-yl)piperidine-1-carboxylate (7). A
solution of 6 1232.14 g (2.7599 mol) and acetic acid (3314 g,
55.2 mol) in methanol (3080 mL) was prepared. A solution of
ammonium acetate (2162 g, 8.94 equiv, 28.05 mol) and
methanol (5600 mL) was prepared separately.
Separate high-pressure syringe pumps were used to flow the
two solutions. The solutions mixed at a standard T-shaped
fitting and then immediately flowed through a stainless steel
coiled tube reactor (221 mL reactor listed in Table 10) which
had been pressurized to 1000 psig with nitrogen and heated in
an oil bath to 140 °C. Flow rates were controlled such that the
molar ratio of ammonium acetate was 10:1 with respect to 6
and Vreactor/Qfeed = 10 min at which point the solution cooled
to ambient temperature in a tube-in-tube heat exchanger
and exited the system via the back-pressure regulator to be
collected. HPLC samples were pulled via an automated
sampling and dilution system every 10 min throughout the
course of the 20 h run. Typical steady-state composition by
HPLC was 16 (3.7%), 1 (0.2%), 17 (0.1%), and 2 (95.5%).
HPLC conditions: Zorbax SB Phenyl 4.6 mm × 50 mm, 1.8 μm;
1.0 mL/min, 40 °C; 260 nm Solvent A: 0.2% (w/v) NH4OAc
in H2O, Solvent B: 0.1% (v/v) AcOH in ACN; Gradient:
0 min (80% A/20% B); 10 min (20% A/80% B); 11 min (20%
A/80% B); 11.1 min (80% A/20% B).
The product solution was divided equally between identical
22 L flasks (6640 g in each), and the following procedure was
followed on each flask to crystallize. The clear, orange solution
was heated to 50 °C under nitrogen. Water (6 L) was added
dropwise at 50 °C. After 2000 mL had been added, the solution
was seeded and the product precipitated. After 6 L was added,
the tan slurry was heated at 50 °C for 1 h; then the heat was
turned off, and the slurry was allowed to cool to room tem-
perature in the heating mantle overnight. The total amount
of water used to crystallize all of the product in both flasks was
12 L. The product was filtered and the cake washed with 1:1
MeOH/water (3 L). The solids were dried in a vacuum oven at
50 °C to afford 7, 1009.5 g (75.3% yield). See earlier experi-
mental for characterization data.
acetate (70 mL). The wet cake was slurried with ethyl acetate
(35 mL) at room temperature, and the mixture was filtered. The
cake was dried under vacuum at 40−50 °C to afford 24 (56.4 g,
1
208.3 mmol, 49% yield from 21) as a white solid. H NMR
(DMSO-d6) δ 9.48 (s, 2 H), 8.36 (m, 2 H), 7.77 (t, 1 H, J =
10.0 Hz), 4.82 (s, 2 H), 2.59 (s, 3 H); 13C NMR (DMSO-d6) δ
91.0, 136.2, 136.1, 131.1, 131.1, 128.3, 121.2, 118.9, 118.7, 54.0,
33.1; IR (KBr mull cm−1) 2922, 2781, 2703, 1687, 1333, 1278,
1146, 1053, 848 cm−1; Exact Mass: Calcd m/z for C10H10ONF4:
236.0693, Found: 236.0221. See Supporting Information for
HPLC chromatogram.
tert-Butyl 4-((2-(4-Fluoro-3-(trifluoromethyl)phenyl)-
2-oxoethyl)(methyl)carbamoyl)piperidine-1-carboxylate
(6). To a 3 L three-neck RBF was added 24 (800 g, 2.95 mol),
1-(tert-butoxycarbonyl)piperidine-4-carboxylic acid (743 g,
3.24 mol), and THF (8.0 L). The reaction mixture was cooled
to 0−5 °C using an ice−water bath. 1-Propanephosphonic acid
cyclic anhydride (T3P, 50% wt. solution in ethyl acetate, 3186 g,
5.02 mol) was added dropwise at a rate sufficient to maintain an
internal temperature of 0−5 °C. Finally, N-methylmorpholine
(3186 g, 11.56 mol) was added dropwise at a rate sufficient
to maintain an internal temperature of 0−5 °C. Upon
N-methylmorpholine addition, the reaction mixture became a
solution. The reaction mixture was warmed to room temper-
ature and was stirred for 20 h. Ethyl acetate (6.4 L) was added,
followed by water (6.4 L). The layers were separated. The
aqueous layer was extracted with ethyl acetate (3.2 L). The
combined organic layer was washed with brine (3.2 L) and
dried over Na2SO4. The mixture was then filtered, and the
filtrate was concentrated under reduced pressure to afford
crude 6 as a dark oil. Ethyl acetate (2000 mL) was added to the
dark oil, and a solution was formed. Heptane (7000 mL) was
added to the solution dropwise, and a slurry was observed. The
resulting slurry was stirred at room temperature for 3 h. The
mixture was then filtered, and the cake was dried under reduced
pressure at 40−50 °C to afford 6 (850 g, 1.90 mol, 65% yield)
4-(4-(4-Fluoro-3-(trifluoromethyl)phenyl)-1-methyl-
1H-imidazol-2-yl)piperidine (5). A solution of 7 (1000 g,
2.34 mol) in a mixture of tetrahydrofuran (20 L) and methanol
(1 L) was flowed through a stainless steel coiled tube reactor
(221 mL reactor listed in Table 10) which had been pressurized
to 1000 psig with nitrogen and heated in a standard GC oven to
270 °C. The flow rate were controlled such that Vreactor/Qfeed
=
15 min at which point the solution cooled to ambient tem-
perature in a tube-in-tube heat exchanger and exited the system
via the back-pressure regulator to be collected. The total
collected reaction mixture was concentrated to a solid (771 g,
97% yield, potency corrected). 717 g of this solid was dissolved
in toluene (4.8 L) in a 22 L RBF at 50 °C. Heptane (7 L) was
added slowly via add’n funnel. After 1 L had been added the
solution was seeded and thin slurry developed. The remaining
heptane was added over ∼1 h and the slurry stirred at 50 °C for
20 min, then allowed to cool to room temperature overnight in
a heating mantle. Solids were filtered, and the cake was washed
with 1 L of 10% toluene in heptane. Solids were dried in a vacuum
oven at 50 °C overnight to afford 5 (573.6 g, 80% recovery) as a
1
as a white solid. H NMR (CHCl3) δ 8.8.22 (d, 1 H, J =
6.8 Hz), 8.16 (m, 1 H), 4.76 (s, 2 H), 4.13 (m, 2 H), 3.17 (s,
3 H), 2.77 (m, 2 H), 1.72 (m, 4 H), 1.44 (s, 9 H); 13C NMR
(CHCl3) δ 192.1, 175.2, 154.7, 134.0, 133.9, 131.5, 131.4,
127.7, 127.7, 117.7, 117.5, 79.6, 54.3, 38.5, 36.6, 28.4, 28.0; IR
(KBr mull) 2981, 2926, 1633, 1505, 1366, 1281, 1215, 1140,
1
white solid (mp 147−150 °C). H NMR (400 MHz, DMSO) δ
1000
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
(2) Katritzky, A. R., Ramsden, C. A., Joule, J. A., Zhdankin, V. V., Eds.
Handbook of Heterocyclic Chemistry, 3rd ed.; Oxford: New York, 2011;
Vol. 133.
(3) Chen, S.-B.; Tan, J.-H.; Ou, T.-M.; Huang, S.-L.; An, L.-K.; Luo,
H.-B.; Li, D.; Gu, L.-Q.; Huang, Z.-S. Bioorg. Med. Chem. Lett. 2011,
21, 1004.
(4) Zotova, S.; Shvedov, V. Pharm. Chem. J. 1994, 28, 103.
(5) Revuelta, J.; Machetti, F.; Cicchi, S. Mod. Heterocycl. Chem. 2011,
2, 809.
7.99−7.96 (m, 2H), 7.60 (s, 1H), 7.42 (t, J = 9.9 Hz, 1H), 3.58 (s,
3H), 2.97 (d, J = 12.3 Hz, 2H), 2.83−2.76 (m, 1H), 2.54 (td, J =
11.9, 2.1 Hz, 2H), 1.70−1.57 (m, 4H); 19F NMR (376.2 MHz,
DMSO-d6) δ −61.48 (d, J = 13.29 Hz, 3F), −111.84 (m, 1F); IR
(neat solid, ν/cm−1) 3304.4, 2939.8, 2805.8, 1555.5, 1495.9,
1458.7, 1436.4, 1324.8, 1295.0. 1228.0, 1198.2, 1116.4, 1049.4,
960.1. 893.1, 833.6, 788.9, 759.1; HRMS (ESI): predicted for
+
C16H18F4N3 : 328.1423, found 328.1431. See Supporting
Information for HPLC chromatogram.
(6) Weintraub, P. M.; Meyer, D. R.; Aiman, C. E. J. Org. Chem. 1980,
45, 4989.
ASSOCIATED CONTENT
(7) It is also interesting to note that resin-bound azide performed
very well in alcoholic solvents as a safer alternative to using solid
sodium azide. Unfortunately, this azide source did not work as well in
THF, which was required to telescope the steps.
(8) Gupton, J. T.; Miller, R. B.; Krumpe, K. E.; Clough, S. C.; Banner,
E. J.; Kanters, R. P. F; Du, K. X.; Keertikar, K. M.; Lauerman, N. E.;
Solano, J. M.; Adams, B. R.; Callahan, D. W.; Little, B. A.; Scharf, A. B.;
Sikorski, J. A. Tetrahedron 2005, 61, 1845.
■
S
* Supporting Information
Select HPLC data as an additional demonstration of purity for
compounds 11a, 13, 24, 6, and 5. This material is available free
of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
(9) This generates very little pressure in the vessel, but the sealed
system prevents ammonia loss and allows for full conversion without
recharging ammonium acetate.
Corresponding Author
*E-mail: (S.A.M.) May_Scott_A@Lilly.com; (J.D.M.)
Johnson_Martin_D@Lilly.com.
(10) As measured from solvent-only runs.
Present Address
(11) This results in feed rates of 0.011 mL/min of a 0.2442 M
solution of ketoamide feed and 0.028 mL/min of a 0.0964 M solution
of TEA/NH4OAc.
(12) For safety, simplicity, and cost reasons, forced convection ovens
are used in place of oil baths for temperatures >140 °C.
(13) Full details of this equipment set are available in the
Experimental Section.
‡D&M Continuous Solutions, 7814 Pennyroyal Lane, Indian-
apolis, IN, 46237.
Notes
The authors declare no competing financial interest.
▼Deceased December 26, 2007.
(14) Reaction composition for this temperature screen was 17 vol
equiv of methanol, 1 equiv of triethylamine, and 10 equiv of
ammonium acetate.
(15) Since mean residence time was not known for each of the
temperatures in the screen, these data could not be used for
understanding the reaction kinetics. This would be important if one
wanted to apply data to other reactor dimensions or types such as
stirred tanks.
(16) We determined that triethylamine was not necessary prior to the
stoichiometry optimization.
(17) This is with a relatively long reaction without any integrated
PAT to aid in sample collection and analysis.
(18) Decreasing d to 0.254 mm, a standard size for stainless steel
tubing, would further reduce the material requirement. This has been
necessary when dealing with new route scouting where material is not
abundantly available. Usually, however, the material availability is
sufficient for using tube reactors with d = 0.56 mm which is less prone
to clogging.
(19) One equivalent of triethylamine was present in the 20-g scale-
up. It was later found that TEA had no effect on the outcome of the
reaction and TEA was not used in the 775-g-scale reaction.
ACKNOWLEDGMENTS
■
We thank Paul Milenbaugh and Ed H. Deweese of D&M
Continuous Solutions (Greenwood, IN) for assembly and
operation of most of the PFR systems in this paper. We thank
the reviewer who recommended improvements to the axial
dispersion sections of this paper. We thank Adam McFarland
and Randy Lambertus for on-line Raman probe measurements
used to characterize the tube reactors for axial dispersion. We
also thank Zheng Chen, John P. Schafer, and John Howell for
select work on the alkylation of compound 2 and scale up
shown in Table 1. We also acknowledge Asymchem for the
pilot-scale preparation of ketoamide 1. Finally we thank Bret
Huff for his vision and support in continuous processing at
Lilly.
REFERENCES
■
(1) Alvarez-Builla, J., Vaquero, J. J, Barluenga, J., Eds. Modern
Heterocyclic Chemistry; 2011; Vol. 2
1001
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002

Organic Process Research & Development
Article
(20) Other oligomeric impurities were present, but 18 was
specifically identified in HPLC traces. 18 can also undergo further
reaction to deliver numerous oligomeric impurities.
(41) Levenspiel, O. Longitudinal Mixing of Fluids Flowing in
Circular Pipes. Ind. Eng. Chem. 1958, 50, 343−346.
(42) Levenspiel, O. The Chemical Reactor Omnibook; Oregon State
University Book Stores, Inc.: Corvallis, OR, 1993.
(43) Table 10 can serve as a guideline for selecting an appropriate
reactor.
(44) Hickman, D. A.; Sobeck, D. D.; In Liquid Phase Process
Characterization. Micro Instrumentation: for High Throughput Exper-
imentation and Process Intensification - A Tool for PAT; Koch, M. V.,
VandenBussche, K. M., Chrisman, R. W., Eds.Wiley, New York, NY,
2007; Chapter 13.
(21) While no significant pressure builds because the 1-butanol
boiling point is 118 °C, the loss of ammonia is an issue in normal
glassware. Running this type of reaction require recharging of
ammonium acetate to drive the reaction to completion.
(22) Five equivalents of NH4OAc at Vreactor/Qfeed = 90 min.
(23) Since we did not observe product degradation, even at elevated
temperatures and for long residence times, we concluded that the
degration of the ketoamide was the primary issue.
(24) Levenspiel, O. Chemical Reaction Engineering. An Introduction to
the Design of Chemical Reactors; John Wiley and Sons, Inc.: New York,
NY, 1962.
(25) A second campaign was later conducted where the reaction,
workup, and isolation were all conducted continuously. This was
beyond the scope of this paper and will be reported in due course.
(26) Grignard reagent was made through exchange between
isopropylmagnesium chloride and 4-fluoro-3-trifluoromethyl-1-bromo-
benzene in THF at RT.
(45) Green, D. W.; Perry, R. H. Perry’s Chemical Engineers’
Handbook.8th ed. McGraw-Hill: New York, 2008.
(46) Bromide, 7a, is also commercially available, but due to project
timelines, we chose to make it in this case.
(47) Typically <7% acetophenone, ∼8% bis-Br, ∼84% 7a.
(48) Despatch Industries, LND1-42-3, Inert Atmosphere Oven for
use in Class I, Division 2 Area.
NOTE ADDED AFTER ASAP PUBLICATION
■
(27) The first experiments showing this result were conducted in
acetic acid as the solvent with a large excess of ammonium acetate.
(28) Klint, D.; Bovin, J.-O. Mater. Res. Bull. 1999, 34, 721.
(29) Reik, R. Monatsh. Chem. 1902, 23, 1033.
This paper was published on the Web on April 19, 2012.
Figure 13 has been updated and the corrected version was
reposted on April 26, 2012.
(30) Ammonium acetate has a solubility of 45.9 mg/mL in methanol
and 16.3 mg/mL in ethanol.
(31) One notable result from this screen, however, is the finding that
an 86% in situ yield can be obtained after just one minute residence
time at 180 °C, even in the presence of 20 equiv of AcOH. This
represents a 1200× rate acceleration vs the batch reaction conditions
at only slightly lower yield.
(32) In the 7 L PFR described in this paper, 44.8 kg of 6 could be
produced in a day. Increasing the reactor to 73 L described elsewhere
by us would afford 467 kg/day throughput.
(33) The yield on 1 kg scale was lower than expected and likely due
to potency issues with the batch of 1 used. Reactions in smaller
reactors were higher, and there is no reason to expect a yield difference
between reactors.
(34) τ was measured in a separate flow experiment using product
only in the reaction solvent.
(35) Not shown because the F-curves for the 221 mL PFR look
practically identical to those for the 4.51 mL PFR.
(36) Global FIA, 684 Sixth Ave. FI, Fox Island, WA 98333, United
States.
(37) It is important to have safety precautions in place for any
unattended operation. Pressure relief for the reactor and pumps and
automatic shutoffs on the pumps are essential in the event the system
becomes clogged and overpressurizes.
(38) (a) Levenspiel, O. The Chemical Reactor Omnibook; Oregon
State University Book Stores, Inc.: Corvallis, OR, 1993. (b) Levenspiel,
O. Chemical Reaction Engineering. An Introduction to the Design of
Chemical Reactors; John Wiley and Sons, Inc.: New York, NY, 1962.
(c) Fogler, H. S. Elements of Chemical Reaction Engineering, 3rd ed.;
Prentice Hall PTR: Upper Saddle River, NJ, 1999. (d ) Walter, J.,
Weber, J., DiGiano, F. A., Eds. Process Dynamics in Environmental
Systems; John Wiley and Sons, Inc.: New York, NY, 1996. (e) Aris, R.
On the Dispersion of a Solute in a Fluid Flowing through a Tube. Proc.
R. Soc. Lond., Ser. A 1956, 235, 67−77. (f) Taylor, G. Dispersion of
Soluble Matter in a Solvent Flowing Slowly through a Tube. Proc. R.
Soc. Lond., Ser. A 1953, 219, 186−203. (g) Taylor, G. Dispersion of
Matter in Turbulent Flow through a Pipe. Proc. R. Soc. Lond., Ser. A
1954, 223, 446−468. (h) Taylor, G. Conditions under which
dispersion of a solute in a stream of solvent can be used to measure
molecular diffusion. Proc. R. Soc. Lond., Ser. A 1954, 223, 473−477.
(39) Levenspiel, O. Chemical Reaction Engineering. In An
Introduction to the Design of Chemical Reactors; John Wiley and Sons,
Inc.: New York, NY, 1962.
(40) Dankwerts, P. V. Ind. Chemist 1954, 3, 102.
1002
dx.doi.org/10.1021/op200351g | Org. Process Res. Dev. 2012, 16, 982−1002